Spatiotemporal Variations in Strain Release and Seismic Rupture in Multifault Systems: An Example from Panamint Valley, Southeastern California Open Access

Geometrically complex, multifault earthquakes are uniquely hazardous to the extent that they involve seismogenic strain transfer across kilometer-scale fault separations or discontinuities. These types of ruptures may be characteristic of immature fault zones [1], such as those occurring within the Eastern California shear zone (ECSZ) (Figure 1(a)). Seismogenic strain transfer across geometrically complex fault networks requires favorable stress conditions, controlled in part by discontinuity-bounding fault geometries, history of prior moment release, and coseismic rupture direction and displacement [2-8]. However, spatiotemporal variations in the boundary conditions that affect fault stresses lead to fault segments that can variably propagate or arrest dynamic rupture over subsequent earthquake cycles (e.g. [9-13]). Identifying how frequently, and under what stress conditions, ruptures can propagate through discontinuous boundaries requires detailed knowledge of fault geometry and information on the persistence or variability of past rupture paths over multiple historical and paleo-earthquake cycles [14]. Documenting this variation involves the generation of paleoseismic rupture maps, for multiple paleoseismic earthquakes along a single fault zone, to quantify the geometry and repeated rupture behavior of surface-rupturing strands.

In this study, we generate paleoseismic rupture maps to document the variations in spatial rupture patterns of a structurally complex and spatially discontinuous fault stepover in the Panamint Valley, in the northern ECSZ (Figures 1(a) and 1(b)). Using these paleo-rupture maps, we assess the spatial extent of past earthquakes and the persistence of surface ruptures in this fault boundary zone that may accommodate multifault rupture. The ECSZ (Figure 1(a)) contains networks of broadly distributed, interconnected, and branching fault systems [1, 12] and collectively accommodates 20%–25% of Pacific-North America plate motions [15-17]. Several geometrically heterogeneous multifault earthquakes have ruptured in the ECSZ within the last ~30 years (Figure 1(a)), including the 1992 Landers earthquake sequence (M_w 6.1 Joshua Tree, M_w 7.3 Landers, and M_w 6.2 Big Bear) [18-20], the 1999 M_w 7.1 Hector Mine earthquake [21, 22], and the 2019 Ridgecrest earthquake sequence [23-25]. Furthermore, evidence has been documented for a paleoseismic multifault rupture on the Calico-Hidalgo fault system south of the Garlock fault, between the Landers and Hector Mine earthquakes [26]. While these historical and prehistoric earthquakes imply that geometrically complex multifault ruptures may be a common rupture style in the ECSZ (Figure 1(a)), there are very few faults where we can document the behavior and/or variability of multifault ruptures over multiple earthquake cycles. This is partly due to either a lack of high-resolution paleoseismic records or a reliance on sparse paleoseismic trenching data, which extrapolates rupture history for a fault based on a single location and may omit spatial variability from rupture histories. Alternatively, paleo-rupture maps can be incredibly useful in identifying changes to the spatial distributions of paleoseismic surface ruptures over multiple earthquake cycles. In this way, paleo-rupture maps can provide a higher spatial resolution of paleoseismic events when compared to paleoseismic trenching, particularly in regions of geometrically complex or distributed faulting.

Here, we leverage new detailed tectono-geomorphic mapping and a locally calibrated, high-resolution late Holocene fan stratigraphy to generate paleoseismic rupture maps that characterize the variations in geometry and location of surface ruptures over four earthquake cycles in the Panamint Valley, ECSZ (Figure 1(b)). We focus on a ~10-km-wide zone of spatially distributed, well-preserved transtensional surface ruptures in late Quaternary alluvial fans located between the Ash Hill and Panamint Valley faults, which we term the Panamint Valley transtensional relay (PVTR, Figure 1(b)). Prior tectono-geomorphic mapping and paleoseismic trenching have demonstrated a similarity in the number and timing of late Holocene earthquakes on the Ash Hill and Panamint Valley faults [27, 28], suggesting that these two faults may experience multifault ruptures via seismogenic strain transfer across the PVTR. To analyze the central Panamint Valley for spatiotemporal variations of surface-rupturing earthquakes, we use our high-resolution alluvial fan stratigraphy, based on semi-quantitative surface morphology and clast weathering metrics (e.g. [28-30]), to subdivide generations of offset alluvial units that bracket the timing of individual earthquakes in the PVTR. We then use feldspar post-infrared infrared-stimulated luminescence (pIR-IRSL) to date and correlate offset alluvial fan deposits and to provide chronologic constraints for individual earthquake ages. This mapping-based approach allows us to correlate geomorphically similar offset deposits over kilometer-scale distances, between fans with different magnitudes of transport and variable proportions of parent lithologies, to map the spatial extent of Holocene earthquakes in the PVTR. Using these methods, we generate paleo-rupture maps to assess how the larger-scale spatial distribution of paleoseismic earthquakes varies over multiple seismic cycles in this region. Our results document spatiotemporal variations in surface-rupturing earthquakes within the PVTR and provide evidence that this fault zone may act as a geometric link to transfer multifault rupture between the Ash Hill and Panamint Valley faults.

The ECSZ (Figure 1(a)) is a network of transtensional and transpressional faults east to southeast of the Sierra Nevada, extending from southwestern Nevada to the Salton Trough [31], that collectively accommodates ~8.5–10 mm/yr of Pacific-North America plate motion [17, 32-36]. The northern ECSZ, also referred to as the southern Walker Lane, is separated from the southern ECSZ by the 250 km, NE-SW trending, left-lateral Garlock fault (Figure 1(a)). In the northern ECSZ, regional strain is distributed on four major right-lateral, strike-slip, and/or oblique-normal faults and fault systems including: (1) the Stateline fault [37], (2) the Fish Lake Valley-Death Valley faults [36], (3) the Saline Valley-Hunter Mountain-Ash Hill-Panamint Valley faults [27, 28, 38-40], and (4) the White Mountain-Owens Valley-Airport Lake faults [41] (Figure 1(a)). These fault zones contain networks of interconnected faults that formed coevally with a reorganization in western North American plate motion in the northern ECSZ between ~11 Ma and ~2 Ma, resulting in a switch from E-W directed extension along low-angle normal faults to NW-SE directed transtension along high-angle, normal, transtensional, and strike-slip faults (e.g. [42-45]). In the Panamint and Searles Valleys, this tectonic reorganization beginning at ~4.6 Ma [39] has led to the progressive abandonment of low-angle, W- to NW-dipping detachment faults (the Emigrant, Panamint, and Slate Range detachment faults; Figure 1(b)), in favor of slip along more optimally oriented, moderate to high-angle, normal, strike-slip, and/or transtensional faults (the Panamint Valley, Ash Hill, PVTR, Manly Pass, Searles Valley, and Tank Canyon faults; Figure 1(b)) [45].

The PVTR is a ~10-km-wide zone of distributed deformation located at the juncture of the Panamint Valley, Ash Hill, and Manly Pass-Searles Valley faults in the central Panamint Valley (Figure 1(b)). The PVTR is located ~4 km southeast of the southern tip of the Ash Hill fault, ~4 km west of the central Panamint Valley fault, and ~2 km northeast of the Manly Pass-Searles Valley fault zone (Figure 1(b)). Scarps within the PVTR do not directly merge with the adjacent Ash Hill or Panamint Valley faults at the surface. However, the subsurface geometries of the PVTR, Ash Hill, and Panamint Valley faults are uncertain, and these systems may directly merge at depth. Below we discuss the existing knowledge of Holocene geometry, slip kinematics, and slip rates along the Panamint Valley, Ash Hill, and Manly Pass-Searles Valley faults. These data provide key boundary conditions for evaluating how strain may be transferred across the PVTR.

The Panamint Valley fault is a ~100 km long, S- to SE-striking (~150–170°), west-dipping, normal to transtensional fault extending between the Garlock and Hunter Mountain faults (Figure 1(a)) and bounds the Panamint Valley on its eastern flank (Figure 1(b)) (e.g. [27, 38-40, 46-51]). The slip kinematics of the Panamint Valley fault vary from dextral-oblique along southern Panamint Valley to normal-oblique slip north of the PVTR (Figure 1(b)) [27, 38-40, 47-51]. In general, the southern Panamint Valley fault is a high-angle, west-dipping transtensional fault with an average slip vector azimuth of ~330° [38, 47], and a strike-slip to dip-slip ratio of ~1.7:1 [38]. While the geometry and slip of the southern Panamint Valley fault are well-documented, the subsurface geometries of the north and central segments of the Panamint Valley fault are debated, and two models have been presented. The first model, the Southern California Earthquake Center (SCEC) Community Fault Model (CFM) version 5.0 [52, 53], suggests that the north and central segments of the Panamint Valley fault are high-angle (>60°) west-dipping normal faults, supported by seismic imaging near Wildrose graben (Figure 1(b)) that shows high-angle normal faults offsetting the low-angle Panamint detachment in the upper ~200–300 m [54]. A second model, proposed by Densmore and Anderson [55] and illustrated by Walker et al. [38], suggests that the dip of the central Panamint Valley fault decreases to <30° north of ~36°N, supported by field observations, geophysical surveys, and palinspastic reconstructions of Panamint Valley [38-40, 45, 50]. In either case, the Panamint Valley fault is the primary structure accommodating late Pleistocene to Holocene deformation in Panamint Valley, with slip rates ranging from 1.5 to 6.0 mm/yr, determined from cosmogenic and luminescence dating of offset geomorphic piercing lines [47-51].

The Ash Hill fault is a ~40 km long, ~SSE-striking (~160–165°), steeply west-dipping (~70–90°), dextral-oblique fault bounding the Panamint Valley on its western flank (Figure 1(b)) [28, 55]. Tectono-geomorphic mapping of late Holocene deposits along the Ash Hill fault documents right-lateral slip with a minor west-side-down normal component, a slip vector azimuth of ~345°, and strike-slip to dip-slip ratios ranging between 5:1 to 8:1 [28, 55]. The Ash Hill fault is the second most active structure in Panamint Valley, with late Pleistocene to Holocene slip rates of 0.6–1.4 mm/yr, determined from cosmogenic dating of Marine Isotope Stage 6 shorelines and luminescence dating of offset Holocene alluvium [28, 50, 55-57]. In accordance with the two end-member geometric models described above, the subsurface fault geometry of the Ash Hill fault is proposed to either be subparallel to a high-angle Panamint Valley fault or merge with a low-angle Panamint Valley fault at depth [28, 55].

South to southeast of the Panamint Valley, the Manly Pass-Searles Valley fault zone (Figure 1(b)) consists of several contemporary subsidiary faults including the Manly Pass, Searles Valley, and Tank Canyon faults (Figures 1(a) and 1(b)) (e.g. [38, 45, 46, 52, 53, 58-60]). The Manly Pass-Searles Valley fault zone is ~45 km long, NW to E striking, shallowly to moderately W-NW dipping (20, 50°), with normal and/or left-lateral displacement [38, 58-60]. These faults are the primary structures bounding the northern and western flanks of the Slate Range, southwest of the PVTR Figure 1(b). While these faults are interpreted to be active, the only existing temporal constraints for slip along the western margin of the Slate Range suggest that the Slate Range detachment accommodated rapid extension at ~4.2 Ma [45], followed by slower late Pleistocene to Holocene slip rates of ~0.17–0.35 mm/yr [60].

In the Panamint Valley, similarities in the number, timing, and recurrence of late Holocene earthquakes have been identified from paleoseismic trenching and tectono-geomorphic mapping along the Panamint Valley and Ash Hill faults [27, 28]. Paleoseismic trenching of the southern Panamint Valley fault provides evidence for at least four late Holocene earthquakes at >4.1 ka, 3.6–3.3 ka, 3.0–0.9 ka, and 0.5–0.3 ka [27], and tectono-geomorphic mapping indicates that late Holocene ruptures may extend to the central Panamint Valley fault at the latitude of the PVTR [50, 51]. Furthermore, detailed tectono-geomorphic mapping has documented at least three late Holocene earthquakes along the southern Ash Hill fault, at 4.3–2.5 ka, 2.5–0.6 ka, and 0.7–0.3 ka [28]. In addition to this temporal overlap, the similarities in fault geometry and slip vectors between the southern Panamint Valley and Ash Hill faults could indicate that the central Panamint Valley is a preferred pathway for paleoseismic multifault or triggered ruptures [28], with seismogenic strain transfer occurring across the PVTR.

Previous interpretations have attributed the faults of the PVTR to strain transfer between the Ash Hill and Manly Pass-Searles Valley fault zone [38]. However, the slower late Pleistocene to Holocene slip rates along the western margin of the Slate Range [60], and the similarities in earthquake timing along the Panamint Valley and Ash Hill faults, suggest that the geometrically complex faults of the PVTR may be more likely a result of the interaction between the relatively faster-slipping Ash Hill and Panamint Valley faults. Additionally, the fault geometries for the Ash Hill and Panamint Valley faults detailed above imply that the Ash Hill fault may sole into an active, low-angle Panamint Valley fault at depth, providing a favorable geometry for strain transfer [28, 38-40, 55]. In this study, we investigate the fault geometries and timing of late Holocene surface-rupturing earthquakes preserved in the PVTR, to identify how strain is accommodated in central Panamint Valley, in the stepover between the Ash Hill and Panamint Valley faults.

We present new surficial mapping of the PVTR over ~46 km² within central Panamint Valley to determine the ages of alluvium that bracket late Holocene earthquakes and to correlate these units and the timing of surface ruptures along fault strike. We completed mapping at two scales: 1:12,000 regional mapping and 1:4,000 high-resolution mapping in selected areas with high surface rupture density. Mapping was completed using 2 m digital surface models (DSMs) [61], 0.5–1 m GeoEye aerial imagery [62], and National Center for Airborne Laser Mapping (NCALM) 0.5 m airborne lidar collected from the EarthScope SoCal Lidar Project [63]. In addition, we collected high-resolution (5.0 cm) structure from motion DSMs in regions with high concentrations of surface ruptures in late Holocene alluvium, or where lidar DEM coverage is absent. These high-resolution DSMs were required to image offsets in small-scale (<1 m wavelength) channels, bars, and swales, which cannot be resolved in existing 0.5 m lidar DEMs.

Mapping and correlating Holocene surface ruptures over several kilometers of fault length requires the development of a locally calibrated, semi-quantitative, high-resolution alluvial stratigraphy. Our surficial mapping, therefore, subdivides generations of alluvial fans based on an extensive set of semi-quantitative morphologic indicators that we detail in the following sections. We combine inset, burial, and onlap relationships, relative elevation above the modern wash, and changes in bar and swale morphology to interpret the relative age of an alluvial surface [64, 65]. These metrics can be used to approximate the age of the underlying deposit in instances where resurfacing and bioturbation are minimal, conditions that are met in the majority of the late Pleistocene to Holocene alluvial units present in the PVTR. Additionally, these metrics can apply to both fluvially transported alluvium and debris flow levees, with minor intra-generational variations of desert pavement development and weathering, dependent on the dominant grain sizes and dominant lithologies of the initial deposit.

We subdivide late Pleistocene to Holocene alluvial fan surfaces into 10 generations following the existing schema of Regalla et al. [28] and Sethanant [51]. This schema combines observations of bar and swale morphostratigraphy and degradation, development of desert pavement (e.g. [66-70]), strength of clast rubification and varnish, variations in the mechanical and/or chemical weathering of clasts [71], and the thickness of the A_v horizon (see Online Supplementary Tables S1 and S2 for specifications of semi-quantitative parameters). These parameters change observably in the first 100s–1000s years following deposition [28, 29, 47, 51], are very well-preserved in arid regions where resurfacing is minor, and act as good relative-age markers for subdividing generations of late Holocene alluvium. To develop this chronology, we produced detailed descriptions of all semi-quantitative parameters along 5–10 randomly selected locations within a single deposit, at dozens of individual sites. Once we had identified unit boundaries and defined descriptions for all our parameter categories, we continued to perform frequent spot checks to identify local or regional variations related to source lithology, transport distance, or transport mechanism. Additionally, detailed surface descriptions were completed for all units at every key rupture site and at sites sampled for geochronology to assess relative unit age. Our convention for naming Quaternary alluvial units involves numbering units in the order in which they were deposited, where the older deposits have smaller numbers (e.g. Units Qf1*, Qf2), and the younger deposits have larger numbers (e.g. Units Qf7, Qf8). This is based on an existing alluvial fan naming schema for the Panamint Valley outlined by McDonald et al. [29] and utilized by Regalla et al. [28].

We use four indicators to characterize the relative age of deposits from bar and swale morphologies: (1) preservation and sharpness of bar and swale boundaries, including length-scale measurements for boundary uncertainty, (2) preservation of depositional imbrication and clast stacking, (3) the proportion of open pockets of air and exposed clast edges, referred to as “open-framework” in bars [28, 29, 72, 73], and (4) percentages of secondary loess infill in the open-framework around surface clasts (Figure 2). These indicators track changes to bar and swale morphology as a function of the potential energy gradient between bars and swales, which causes coarse-grained sediments in bar crests to naturally move downgradient into swales over time. This process leads to an increase in bar-swale boundary uncertainty, and a decrease in clast stacking, imbrication, open-framework, and overall bar relief, further assisted by the infilling of open-framework by eolian fines (Figure 2) [74].

We use four parameters to define the degree of aggradational desert pavement development in gravel-sized clasts: (1) the percentage of clast coverage, (2) interlocking or edge alignment frequency, (3) the strength of clast embedding, and (4) average clast size in bars and swales, independently (Figure 2, Online Supplementary Tables S1 and S2) [75]. Aggradational desert pavements have been documented in the ECSZ as Quaternary surfaces that form when eolian influx accumulates below a surficial layer of alluvium, as the surficial layer of alluvium breaks down mechanically and chemically into smaller gravel-sized clasts [66-70]. The development of desert pavements is strongly reliant on the time-dependent mechanical and chemical weathering of surface clasts, the transport and abrasion of surface sediment by wind, the development of desert varnish and rubification, and the accumulation of iron and pedogenic carbonate in the underlying soil horizons (e.g. [66-70]). We use these four desert pavement development parameters to track the resulting product of these environmental processes through time.

We characterize the relative degree of clast weathering by documenting: (1) the frequency and strength of intra-clast fracturing (Figure 3(a)), (2) variations in the strength of chemical weathering of key indicator minerals such as micas, feldspars, and quartz, (3) the strength of surface relief on clasts associated with the disaggregation and grussification of granites (Figure 3(b), and (4) the strength of surface relief via dissolution (i.e. pitting and etching) of carbonates (Figure 3(c), Online Supplementary Tables S1 and S2). Additionally, we use the abundance of “ghost clasts”(Figure 3(d)), defined as clasts that have been strongly disaggregated at the surface to a conical, rounded, or flat shape, as a key indicator of fan age.

We describe the qualitative hues of varnish and rubification for granitic, basaltic, and carbonate clasts separately, as desert varnish and clast rubification are strongly lithology-dependent. Desert varnish and rubification of surface clasts have been used extensively for correlation of late Quaternary alluvial surfaces in the southwestern U.S. (e.g. [76, 77]) due to the widespread formation and preservation of Mn and Fe oxides and clay coatings on clasts in arid environments that strengthen with surface age. Desert varnish forms through the deposition of Mn and Fe oxides and mixed-layer clays on clast surfaces by wind or water transport [78]. Clast rubification forms as iron precipitates on the undersides of clasts, enhanced by frequent wetting, drying, and oxidation in arid environments associated with ephemeral runoff [79]. Additionally, we document the frequency of “flipped” clasts, defined by subaerial exposure of the basal, rubified clast surface [28, 80]. For “flipped” clasts, we note the strength of revarnishing of the subaerially exposed, rubified surface by assessing the proportion of surface area where varnish has obscured rubification, which is often strongest along the clast edges.

We report the thickness of the upper loess and vesicular (A_v) soil horizons, defined as the layer of wind-blown silts that accumulate immediately beneath the uppermost pavemented alluvial surface. The A_v horizon strengthens systematically with time, appearing thin and patchy in late Holocene alluvium, increasing significantly in thickness, lateral continuity, and degree of cementation in late Pleistocene-aged alluvium [81]. The accumulation and cementation of the A_v layer are strongly time-dependent, making them useful indicators in determining relative surface age, especially when used in conjunction with other parameters described in this study. We focus on measurements of the A_v layer within the soil profile because prior mapping in the Panamint Valley has shown a systematic increase in A_v thickness with age in Holocene fans and because this measurement can be made with minimal disturbance to the surrounding fan surface. We document A_v thickness in pits dug for geochronology sampling, from existing channel cuts, and from dozens of small (<15 cm) pits dug into the top of fan surfaces where fan surface morphology was also described.

We collected drone surveys in targeted, sparsely vegetated regions that lacked existing lidar coverage, and in regions where small magnitude (0.5, 1 m) total offsets in young alluvium are poorly resolved by existing 0.5 m NCALM lidar DEMs (see Online Supplementary Figure S3 for survey locations). This imagery was collected using a DJI Phantom IV drone, receiving signals from GPS and GLONASS for positioning, with a 12.4-megapixel camera. We programmed surveys using DroneDeploy to collect imagery with 75% side-lap and 80% front-lap, at a flight speed of 6 m/s, a starting height of 75 m from the takeoff point, and a maximum flight height of 102 m. We collected 20 surveys, each with an area of 0.16 km², containing 6–8 targets of known dimensions that were surveyed using an Emlid Reach RS RTK differential GPS. We processed the photogrammetry using Agisoft Metashape, following the methods described in Reitman et al. [82], and the U.S. Geological Survey Uncrewed Aircraft Systems Data Post-Processing Guide [83]. The horizontal and vertical resolutions of the orthomosaics and DSMs produced using this method were 2–3 cm/px and 5–6 cm/px, respectively. Resultant DSMs have a horizontal uncertainty of 5 cm (+1.7/-0.5 cm) per meter of horizontal distance, and a vertical uncertainty of 6–10 cm.

Fault traces were mapped from a combination of observations including: (1) sharp topographic steps, (2) a linear alignment of vegetation, (3) a sharp change in the average grain size of desert pavement along a linear trend equivalent to the orientation of other nearby faults, (4) a sharp, high-angle (~70°–90°) diversion in bar and swale flow paths in younger deposits to avoid a degraded scarp, and/or (5) rotated surface clasts and/or a linear disruption of pavement accompanied by an accumulation of loess. Additionally, we mapped fault traces and located offset geomorphic piercing lines using DSM derivatives such as slope, standard deviation, curvature, and topographic profile index (TPI) [84](Online Supplementary Figure S4) . Strike-slip and normal faults are easily identified on DSM derivatives as linear to curvilinear features of high slope, high curvature, high standard deviations of elevation, and where locally sharp steps occur between regions of similar average elevations [85] (Online Supplementary Figure S4). Slope and curvature grids were calculated based on nearest-neighbor approaches, while standard deviation grids were calculated over 0.05–7 m radii. We used two annuli scales for TPI analyses to identify offsets in short- and long-wavelength topography. The fine-scale annulus neighborhood consisted of an inner and outer radius of 1.5 m and 4.5 m, respectively, and permitted the identification of offset short-wavelength and small-amplitude bar and swale morphology. The coarse-scale annulus neighborhood consisted of an inner and outer radius of 15 m and 45 m, respectively, and highlighted ruptures that offset the entirety of an alluvial fan.

We measured lateral and vertical offsets using piercing lines defined by offset bar edges and crests, swale thalwegs, and terrace risers, using a combination of field measurements, topographic profiles, and LaDiCaoz backslipped DSMs [86]. We measured lateral offsets by tracing a range of lateral projections of a piercing line into a fault to estimate the minimum, maximum, mean, and standard deviation of lateral displacement. We measured vertical separation and uncertainty from topographic profiles extracted from our 0.5 cm DSMs using a Monte Carlo simulation (seeOnline Supplementary Material S5 for topographic profile reconstruction methods) that accounts for uncertainty in ground surface slope and fault intersection position, following the methods of Morell et al. [87] and Thompson et al. [88]. We assigned a confidence rating to each offset measurement on a scale from 1 (high confidence) to 3 (low confidence) using measurement quality descriptors to estimate the accuracy and precision of the reported offset, following the approach of Sieh and Jahns [89] and Haddon et al. [90]. For example, a high-confidence offset measurement has a well-defined piercing line that intersects nearly orthogonally with the fault, well-preserved piercing points, and minimal modification, alteration, and/or bioturbation of the surface.

We reconstructed the total oblique slip of offset piercing lines using the lateral and total displacement calculator (LaDiCaoz_v2) algorithm, following the methods of Zielke et al. [86] and Haddon et al. [90]. This approach “backslips” displaced geomorphic features and determines a best-fit probability distribution function (PDF) for fault offset by cross-correlating topography on opposite sides of a fault using relief, width, and degree of symmetry. We applied this method using our newly generated 0.5 cm DSMs using a 0.5 m swath around each fault-parallel profile, positioned within a distance of 1–10 m from the fault trace. For each identified piercing point, we repeated the backslip function between 5 and 8 times to account for uncertainty in the reconstruction of the geomorphic feature. We averaged all repeated measurements of right-lateral and total offset for each offset piercing line to get a single PDF curve (see Online Supplementary Material S6 for our methods of combining PDFs). We report the average and 1σ of this combined PDF. For all other PDFs (e.g. the average displacement of an alluvial unit at key rupture sites), we report the average and 1σ determined using the root sum of squares method.

We then summed averaged PDFs from each offset geomorphic feature to produce cumulative offset probability distributions (COPDs) of fault slip. Clusters of offsets create peaks in the COPD plots that are commonly used to interpret the number and magnitude of slip events along a fault zone [90-92]. Our COPD analysis involved the application of filtering and optimization analyses to determine statistically significant peaks that may represent the cumulative offset from multiple earthquakes. For this analysis, we first summed all confidence-scaled PDFs and removed PDF curves with statistically high uncertainty and/or dispersion (Online Supplementary Material S7(1)). Second, we conducted optimization modeling in RStudio to identify modal peaks that minimized the root mean square error (RMSE) misfit between the observed COPD data and a series of theoretical unimodal, bimodal, or trimodal offset distributions (Online Supplementary Material S7(2)). Finally, for all optimized model curves with 2 or 3 peaks, we performed a 2-sample t-test between peaks to identify if the means are statistically significant to 5%.

We combine our detailed surface rupture mapping in late Pleistocene to Holocene alluvium, with measurements of cumulative offset, and the identification of earthquake-bracketing alluvial generations, to develop maps of paleo-surface rupture networks. We make these maps to identify the patterns of along-strike strain release (i.e. the presence or absence of rupture of a known age) and to identify any variations in the dominant structures that are accommodating coseismic slip over multiple earthquake cycles. To produce paleo-surface rupture maps, we use spatial bins of various sizes for rupture relationships, rather than reporting rupture relationships for each individual surface rupture strand. We use spatial binning of earthquake rupture relationships for two reasons: First, this approach allows us to account for spatial variations in the uncertainty of earthquake timing along the length of each surface rupture strand. For example, paleo-rupture map patterns are best defined where late Holocene earthquakes offset late Holocene alluvium, particularly where there is an increase in cumulative total offset in sequential alluvial generations of increasing age. Paleo-rupture map patterns are more uncertain where late Holocene ruptures offset late Pleistocene alluvium, where one or more earthquake-bracketing late Holocene alluvial units are absent. Second, this approach allows us to relate the deformation mapped at the surface to subsurface slip distributions, assuming that the numerous, closely spaced, transtensional surface ruptures likely merge onto fewer, common structures at depth (e.g. flower structures) [93-95]. In this way, spatial binning of closely spaced surface ruptures yields maps that allow us to interpret potential locations of these common structures within the PVTR that may have ruptured at depth.

We generate separate paleo-rupture maps for each identified late Holocene earthquake, using spatial bins with areas of 0.04 km², 0.16 km², and 0.36 km², based on average and maximum surface rupture spacings ranging between 0.2 and 0.6 km in the PVTR (see Section 4.3.2). We subdivide the area of the PVTR into square bins that are aligned along a NNE-SSW trend, consistent with the average trend of surface rupture strands (Section 4.3.2 ). Each bin is assigned one of three categories: (a) “rupture,” (b) “no rupture,” and (c) “potential rupture.” We assign “rupture” to bins containing one or more surface rupture strands whose earthquake timing is well-constrained, either by increasing magnitudes of offset along strike in sequential generations of alluvium, or by rupture truncation in younger deposits. We assign “no rupture” to bins where none of the surface rupture strands have evidence for rupture in a given earthquake. This includes regions where either tightly constrained rupture brackets indicate a younger bracketing deposit did not experience a rupture, or where we did not identify surface rupture strands from high-resolution mapping. Finally, we assign “potential rupture” to regions where the timing of an earthquake is uncertain. These include regions where we are unable to bracket the minimum timing of an event with an inset younger deposit, or where Holocene ruptures offset Pleistocene-aged alluvium (Section 4.3.1).

We used luminescence dating to further constrain ages for key alluvial and mixed alluvial-spring deposits. For sample dating, we used K-feldspar pIR-IRSL recovered from fine-grained sands in alluvial fans. IRSL has been successfully used to date late Pleistocene and Holocene alluvium in the arid Mojave region where organic material used for C¹⁴ dating is scarce [28, 96, 97] and has provided more reliable age interpretations than optically stimulated luminescence in the Mojave [97-99]. Luminescence dating works by taking advantage of crystal lattice defects that act as a “trap” for electrons, excited from ground states by background ionizing radiation, in the absence of sufficient resetting heat or sunlight (e.g. [99-102]). The accumulated energy is proportional to time by the equation: Age = D_e/D_r, where D_e is equal to the equivalent dose and D_r is the environmental dose rate. The equivalent dose (D_e) is the amount of radiation required to produce the amount of trapped charge, measured in K-feldspar by the number of photons emitted by the mineral after exposure to infrared light. The environmental dose rate is the background radiation produced primarily from the decay of U, Th, and K present in alluvium. Traditional IRSL can display “anomalous fading,” where electrons from higher energy traps transfer to lower energy traps under ambient temperatures, possibly related to quantum tunneling [103-105]. Post-IR-IRSL uses a multi-step infrared stimulation process to remove the anomalous fading signal and find the best electron trap energy that produces the most reliable luminescence ages. For additional details on how D_e and D_r were determined, and to see additional IRSL processing specifications, see Online Supplementary Material S8 and Online Supplementary Tables S9.

We sampled five alluvial fan deposits and one mixed alluvial-spring deposit to further constrain the ages of our relative-age fan chronology. To minimize inheritance, we targeted deposits that displayed evidence for fluvial transport (e.g. well-sorted, presence of bedding or clast imbrication). Within fluvial deposits, we sampled fine to very fine sands (63, 250 µm) which produce the best luminescence signal sensitivity, intensity, stability, and reproducibility [99]. We completed a detailed description of the alluvial surface, soil development, and unit stratigraphy at each sampling location. We evaluated each sample site for evidence of partial bleaching, post-depositional mixing/bioturbation, and evidence of post-depositional high moisture content, as these factors can produce artificially old or artificially young luminescence ages.

We dug 65–86 cm-deep soil pits and produced detailed stratigraphic sections to assess deposition and post-deposition processes, clay content, grain size, sedimentary structures, and soil development. During night sampling, we scraped pit walls back ~3–4 cm to discard surficial (light-exposed) sediment, sieved bulk sediment from targeted layers to remove grain sizes larger than 850 µm and tightly packed and sealed the sediment into 3 cm-diameter opaque steel tubes. We day-sampled one mixed alluvial-spring deposit under an opaque black sheet by digging into a wash cut to expose a fresh soil profile and hammered a metal tube into the exposed deposit wall.

We identify five generations of Holocene alluvial deposits of increasing relative age, based on their surficial and soil characteristics, and their relative landscape position (Figure 4, Table 1); Units Qf6b, Qf6a, Qf7, Qf8, and Qa, from oldest to youngest. See Online Supplemental plate S10 for the full preliminary map plate, list of map units, and map unit correlations. The youngest generation of mapped alluvium, active wash deposits (Unit Qa), is uncemented, loose sands and gravels that have no soil formation, unmodified bar and swale morphology, an absence of desert pavement, and a lack of varnish, rubification, and in-situ clast weathering. In comparison, inactive, late to middle Holocene alluvial deposits, Units Qf6b, Qf6a, Qf7, and Qf8, from oldest to youngest (Figure 5), have weakly to moderately modified bar and swale morphologies, proto- to moderately developed desert pavement, light-tan to dark-brown varnish, yellow to medium-orange rubification, and rare to infrequent, weak to moderate in-situ clast weathering. We map undivided Units Qf6b and Qf6a deposits as Unit Qf6.

Unit Qf8 surfaces (Figure 5, Table 1) are distinguished from younger Unit Qa surfaces by very weakly cemented soils, an increase in abundance of interclast framework fill, very weakly developed varnish and rubification, and small bar margin boundary uncertainties (~10–15 cm). Compared to younger Unit Qa surfaces, Unit Qf8 surfaces also have weak development of proto-pavements, a larger proportion of gravel in swales, and rare, weakly fractured in-situ clasts that are absent in younger Unit Qa surfaces.

Unit Qf7 surfaces (Figure 5, Table 1) are distinguished from younger Unit Qf8 surfaces by weakly cemented soils, an increase in abundance of interclast fill, an increase in clast alignment, and a larger proportion of gravel coverage in swales. Unit Qf7 surfaces have weakly modified bar and swale morphology with small to moderate bar margin boundary uncertainties (~15–25 cm), infrequent to rare stacked and/or imbricated clasts, proto-pavements with common to abundant clast coverage, very weak embedding, and infrequent to rare alignment of clast edges. Compared to younger Unit Qf8 surfaces, Unit Qf7 surfaces have stronger in-situ clast weathering, a more pronounced light-brown varnish, and stronger yellowish-orange rubification.

Unit Qf6a surfaces (Figure 5, Table 1) are generally distinguished from younger Unit Qf7 surfaces by moderately modified bar and swale morphologies, larger bar-margin boundary uncertainties (~25–40 cm), a stronger medium-brown varnish, stronger orange rubification, incipient desert pavement development, rare open-framework, and rare to absent clast imbrication and/or stacking. Unit Qf6a surfaces are the youngest alluvial generation to display incipient desert pavement (Figure 5), with swales displaying >95% gravel coverage, infrequent to common clast edge alignment, and weak (1, 2 cm) embedding of medium gravels.

Unit Qf6b surfaces (Figure 5, Table 1) are distinguished from younger Qf6a surfaces by moderately modified bar and swale morphology, larger bar margin boundary uncertainties (~40–60 cm), stronger desert development with increased embedding strength and cementation, stronger medium- to dark-brown varnish, stronger and more abundant in-situ clast weathering, and more abundant flipped clasts. Unit Qf6b surfaces generally have rare to absent interclast framework, moderately developed desert pavement with weak to moderate clast embedding, and common in-situ clast fracturing.

We identify five generations of late Pleistocene to early Holocene alluvium (Figure 4, Table 1; Qf1*, Qf2, Qf3, Qf4, Qf5, from oldest to youngest), with very well-developed desert pavement containing well-sorted, strongly embedded, and abundantly interlocking fine to medium gravels. In general, we differentiate generations of late Pleistocene to early Holocene surfaces using inset and onlap relationships, the degree of basaltic clast weathering, increasing reddening and metallic hues of rubification and varnish, and increasing accumulation of pedogenic carbonate and iron in soils.

Unit Qf5 surfaces (Figure 5, Table 1), the oldest generation of Holocene alluvium, have rare, relict bar and swale morphology, well-developed desert pavement, strong varnish and rubification on granitic and basaltic clasts, weakly developed varnish and rubification on carbonate clasts, and moderately revarnished, commonly flipped clasts. Unit Qf5 surfaces are distinguished from younger Unit Qf6b surfaces by the presence of ghost clasts and lack of remnant bar and swale morphology.

Unit Qf4 surfaces (Table 1 have very well-developed desert pavement containing well-sorted gravels, with a strong metallic rubification on granitic and basaltic clasts, weakly developed varnish with moderately to strongly rubified carbonate clasts, and commonly flipped clasts with moderately revarnished edges. Unit Qf4 surfaces are distinguished from younger Unit Qf5 surfaces by a greater abundance of ghost clasts, metallic shades of rubification, and strongly weathered carbonate clasts.

Unit Qf3 surfaces Table 1 have very well-developed desert pavement containing well-sorted gravels, with strong metallic-rust-red to rust-black hues of varnish and rubification on all clast lithologies. Qf3 surfaces are distinguished from younger Qf4 surfaces by very strongly weathered quartz in granitic clasts, infrequent revarnishing of flipped clasts to metallic red-brown hues, the inability to distinguish individual lithologies among >90% of clasts, and the weakly to moderately developed desert pavement over ghost clasts.

Unit Qf2 surfaces (Table 1) have very well-developed desert pavement with very strongly embedded and very strongly disaggregated clasts, with strongly developed dark-rust-red to metallic-red-black hues on >90% of surface clasts. Qf2 surfaces are distinguished from younger

Qf3 surfaces by the inability to remove gravel-sized surface clasts from the surface without a chisel, due to the strongly accumulated pedogenic carbonate and deeply embedded clasts. Additionally, Qf2 soils have strong red-orange B horizons, best observed in stream cuts and along fan edges.

Unit Qf1* surfaces (Table 1) have strongly modified desert pavements, an abundance of medium gravels, rare to absent boulders, moderately to strongly embedded clasts, and rare to infrequent ghost clasts. Unit Qf1* has surfaces at higher elevations than younger Qf2 surfaces and can be distinguished from Qf2 by a greater abundance of medium-grained gravels, more weakly developed clast embedding, and thinner A_v soil layers, suggesting some degree of resurfacing.

We identify two generations of lake and spring deposits, Units Qol and Qyl, based on their stratigraphic elevation above modern channels, alluvial burial relationships, and lithification state. Younger, Unit Qyl lake, and spring deposits are consolidated, massive to micritic white limestone, conglomerate, and sandstone with graded bedding that occur stratigraphically below Unit Qf5. Unit Qyl spring deposits are white to cream-colored, unconsolidated, chalky, and produce spongy, white to cream-colored soils. Unit Qol carbonate deposits are 10–20 m thick and outcrop as large, hundred-meter-scale white to cream-colored, massive to micritic limestones with honeycomb/tafoni weathering and occur stratigraphically below Qf2. Unit Qol also outcrops as 10–12 meter-scale, strongly pitted and/or etched carbonate windows where Unit Qf2 deposits are thin.

We determine six new feldspar pIR-IRSL ages, which we use to constrain the chronologic ages of alluvium that we map using our relative-age alluvial fan chronology. We dated one mixed alluvial-spring deposit and five alluvium deposits from pits excavated at four locations, geochronology sites A, B, C, and D (Table 2, Figures 4, 6, and 7). Below we discuss the age constraints of dated deposits, including Units Qf5, Qf6b, Qf6a, Qf7, and Qf8, from oldest to youngest. Unit Qf5, a late Pleistocene to early Holocene deposit, was dated with a single pIR-IRSL sample, PAN2109, at geochronology site A (Figures 6 and 7). This sample was collected from a ~1-m-deep wash cut, at the base of a Qf5 deposit, overlying basal spring deposits (Figure 7, Table 2). The dated material was extracted at a depth ~50 cm from a clast-supported, moderately sorted coarse gravel and sand layer. The deposit yielded a pIR-IRSL age of 20.6 ± 2.9 ka (23.5, 17.7 ka, Table 2), which we interpret as representative of the maximum age of unit Qf5.

Unit Qf6b was dated using three pIR-IRSL samples at geochronology sites B, C, and D, referenced as samples PAN2105, PAN2108, and PAN2104 Figures 4, 6, and 7 and Table 2. Sample PAN2105 was collected at geochronology site B (Figures 4 and 6) from an 86-cm-deep pit that was excavated from a cut bank incised into a Qf6b deposit (Figure 7, Table 2). The dated material was extracted at a depth of 22 cm from a clast-supported, moderately sorted gravel and sand layer. Sample PAN2105 yielded a pIR-IRSL age of 5.38 ± 0.51 ka (5.89, 4.87 ka; Table 2). Sample PAN2108 was collected at geochronology site C (Figures 4 and 6) from a 70-cm-deep pit in a Qf6b deposit (Figure 7, Table 2). The dated material was extracted at a depth of ~44 cm from a moderately stratified layer of medium to fine gravel and coarse sand. Sample PAN2108 yielded a pIR-IRSL age of 4.37 ± 0.23 ka (4.60, 4.14 ka; Table 2). Sample PAN2104 was collected at geochronology site D (Figures 4 and 6) from a 67 cm-deep soil pit dug into a Qf6b deposit (Figure 7, Table 2). The dated material was extracted at a depth of ~26 cm from a moderately sorted, weakly stratified coarse to medium sand deposit. Sample PAN2104 yielded a pIR-IRSL age of 4.20 ± 0.35 ka (4.55, 3.85 ka; Table 2). We use these three pIR-IRSL ages to determine the depositional age of Qf6b to be 5.89–3.85 ka. This Qf6b age is consistent with stratigraphic constraints that require Qf6b to be older than Qf6a (maximum age of ~3.9 ka, see below).

Unit Qf6a was dated using a single pIR-IRSL sample from geochronology site D, referenced as sample PAN2103 (Figures 4 and 6). Sample PAN2103 was collected from a 65-cm-deep pit dug into a Qf6a deposit at geochronology site D (Figure 7, Table 2). The dated material was extracted at a depth of 38–46 cm from a well-sorted fine-grained sand and silt layer. Sample PAN2103 yielded a pIR-IRSL age of 3.58 ± 0.29 ka (3.87, 3.29 ka, Table 2), which we interpret as the depositional age of unit Qf6a.

Unit Qf7 was dated from a single pIR-IRSL sample at geochronology site C, referenced as sample PAN2107 (Figures 4 and 6). This sample was collected from a 72-cm-deep pit excavated from a cut bank incised into a Qf7 deposit (Figure 7, Table 2). The dated material was extracted at a depth of ~26 cm from a clast-supported, moderately to well-sorted cobble layer that fined upward into coarse to medium sand. This deposit yielded a pIR-IRSL age of 2.36 ± 0.20 ka (2.56, 2.16 ka, Table 2), which we interpret as the depositional age of unit Qf7. This age is consistent with stratigraphic constraints that require Qf7 to be younger than Qf6a (minimum age of ~3.3 ka, see above).

Unit Qf8, the youngest generation of deformed late Holocene alluvium, was unable to be directly dated with pIR-IRSL in the PVTR due to a lack of suitable deposits. However, the age of Unit Qf8 can be estimated using two independent constraints. First, Qf8 consistently onlaps Qf7 and must be younger than the 2.56–2.16 ka deposition age of Qf7. Second, Qf8 surfaces mapped along the PVTR have the same morphologic characteristics as deposits mapped along the east side of Panamint Valley, which have a ¹⁴C age of ~0.8–0.5 ka [48, 50, 51]. We interpret the depositional age of Qf7 as a maximum age of Qf8, and the ¹⁴C age, from morphologically similar deposits along the east side of Panamint Valley, as the minimum age of Qf8. These data provide an approximate age bracket of ~2.2–0.5 ka for Qf8 in the PVTR.

Scarps in late Pleistocene and Holocene deposits in the PVTR preserve right-lateral strike-slip and normal slip with east-side-down and west-side-down displacement (Figure 4(b). Transtensional grabens are very common in the PVTR and locally produce enough vertical separation to distinguish between single-event and multi-event scarps, and between recently active and abandoned scarps. We use a combination of field observations and measurements of offset geomorphic features from topographic profiles to semi-quantitatively describe three main categories of scarp morphology including: (1) scarps in Holocene alluvium formed by late Holocene surface ruptures, (2) scarps in late Pleistocene alluvium formed by late Holocene surface ruptures, and (3) older, degraded scarps in late Pleistocene alluvium formed by late Pleistocene surface ruptures that have not been reactivated in Holocene events. These observations were used as an additional line of evidence to document how the spatial distribution of earthquakes varies over multiple earthquake cycles, which is useful for paleo-rupture map interpretations. The identifying features described below may be poorly preserved or harder to distinguish in the absence of a graben, where the slip mode is purely lateral [106], or where slip dies out at a fault tip.

Within scarp morphology category 1, encompassing Holocene scarps in Holocene alluvium, we identify four sub-categories: (1a) young, single-event scarps, (1b) older, degraded, single-event scarps, (1c) older, compound, multi-event scarps that have been reactivated by a younger surface rupture, and (1d) abandoned, compound, multi-event scarps that have not been reactivated by the youngest Holocene earthquakes. Young, single-event scarps in Holocene deposits (scarp morphology (1a)] have average scarp slopes of 8° (± 3°) over horizontal distances of < ~1 m, average vertical separations of 10 cm (± 5 cm) and have unvarnished to very weakly varnished clasts on scarp faces (Figures 8(a) and 8(b)). Older, degraded, single-event scarps in Holocene deposits (scarp morphology 1b) have average scarp slopes of 6° (± 1°) over horizontal distances of 1–3 m, average vertical separations of 10 cm (+6/−4 cm), and have moderately to strongly revarnished clasts on scarp faces (Figure 8(c)). Reactivated, compound, multi-event scarps in Holocene deposits (scarp morphology 1c) have average scarp slopes of 8° (± 3°) over horizontal distances of 1–3 m, average vertical separations of 20 cm (+12/-6 cm), and a mixture of well-varnished clasts and unvarnished clasts on the scarp face. Some reactivated, compound, multi-event Holocene scarps (scarp morphology 1c) can be identified by a sharp break superimposed onto a longer-wavelength, more degraded, pavemented scarp. For example, we observed one scarp with two distinct slopes of 5° and 8°, separated by a distinct submeter step in the scarp face. Additionally, scarps identified as morphology 1c may have cobbles and boulders with noticeable varnish rings (Figure 8(d)). These varnish rings mark sharp boundaries between a varnished clast surface and a more recently exposed, unvarnished, underside of a clast on a scarp, indicating the original position of the stable ground surface prior to earthquake displacement [107, 108]. Abandoned, compound, multi-event scarps in Holocene deposits (scarp morphology 1d) have average slopes of 4° (±1°) over horizontal distances of 1–6 m, average vertical separations of 10 cm (+10/-5 cm), uniform medium- to dark-brown varnish on scarp clasts, and may have developed proto- to incipient desert pavement on scarp faces.

Within scarp morphology category 2, encompassing Holocene scarps in late Pleistocene alluvium, we identify two sub-categories: (2a) young single-event scarps in late Pleistocene deposits and (2b) compound, multi-event scarps in late Pleistocene deposits, where recent surface rupture has reactivated an older late Pleistocene scarp. Young, single-event scarps in late Pleistocene deposits appear as a sharp line of unvarnished to very weakly varnished clasts within a strongly modified, very well-pavemented surface (Figure 8(e)), often with very small vertical separations (~0–5 cm) over a very small horizontal distance (<<1 m). Recently reactivated, compound, multi-event scarps in late Pleistocene deposits (Figure 8(f)) may have one or more 20° (± 4°) slope breaks over 0.3–5 m horizontal distances superimposed on a gentler, longer-wavelength scarp face with slopes of < 10°, over horizontal distances of 5–15 m. These sharp slope breaks may be step-like with vertical separations of ~10–20 cm and are easily identifiable within scarps of larger vertical separations of 5–10 m. These younger scarps often have well-defined sections of unvarnished or very lightly varnished clasts, or strongly varnished clasts with varnish rings (Figure 8(d)), within an otherwise strongly varnished and well-pavemented taller scarp.

Scarp morphology category 3, encompassing late Pleistocene scarps formed in late Pleistocene alluvium, includes strongly degraded, compound, multi-event scarps in late Pleistocene fans with average slopes of 6° (± 3°) over horizontal distances of ~30–100 m, average vertical separations of 4.5 m (+6/−3 m), and very well-pavemented scarp faces with strongly varnished clasts. The pavement development and surface clast varnishes resemble the undeformed late Pleistocene fan surface. Notably, we observed that grabens in late Pleistocene units are often bounded by one abandoned, strongly degraded, multi-event scarp (scarp morphology 3), and one recently reactivated, older, multi-event scarp (scarp morphology 2b).

We group adjacent PVTR surface rupture strands with similar locations and strikes into separate fault sets. We use this strategy to identify patterns in groups of surface ruptures that may coalesce onto common fault planes at depth. This strategy yielded three distinct fault sets (FS 1–3, with each fault set comprised of recent, late Holocene surface rupture traces and older late Pleistocene surface rupture traces (Figures 4(a) and 4(b)). Where Holocene surface rupture strands offset middle to late Holocene deposits (scarp morphology 1, Figures 8(a) through 8(e)), rupture traces occur as ~17–20 subparallel to en échelon strands, with spacings of 1s to 10s of meters. In locations where Holocene surface rupture strands offset late Pleistocene deposits (scarp morphology 2, (Figure 8(f)), individual rupture traces coalesce into sets of ~7–12 subparallel to en échelon strands with spacing of 10s to 100s of meters. Late Pleistocene scarps in Pleistocene deposits (scarp morphology 3) occur as 2–4 subparallel to en échelon strands, with spacings of 0.1–0.5 km. Within each fault set, surface rupture stepovers are 10s to 100s of meters wide and accommodate transtension through a series of grabens and half grabens.

Each set of faults (Figure 4(a)) is ~6–7 km long and is comprised of 10s–100s of surface rupture strands with lengths of 10s–1000s of meters. Fault set 1 (FS 1, Figure 4(a)), the westernmost grouping of surface ruptures, is ~6.3 km long, comprised of surface rupture strands spaced ~30–600 m apart, with an average strike of 182° (+12°/-9°). Within fault set 1, the region south of 36° 2' 30" N contains the highest concentration of inactive scarps in the PVTR, with many of the surface ruptures (~30–70%) consisting of older, inactive, degraded late Pleistocene scarps (scarp morphology 3). In comparison, within fault set 1, north of 36° 2' 30" N, the majority of scarps (~90%) in late Pleistocene deposits have been reoccupied by at least one hrolocene event (scarp morphology 2). Fault set 2 (FS 2, Figures 4(a) and 4(b)), the central grouping of faults in the PVTR, is ~6.6 km long and contains surface ruptures spaced ~3–200 m apart with average strikes of 189° (+30°/−21°). Within fault set 2, the highest concentrations of older, inactive, degraded late Pleistocene ruptures (scarp morphology 3) occur between 36° 0’ 45” N and 36° 1’ 30” N, along the border of fault set 1 and fault set 2 (Figure 4(b)). Elsewhere in fault set 2, the majority (>90%) of scarps in late Pleistocene deposits have been reoccupied by at least one hrolocene event (scarp morphology 2). Fault set 3 (FS 3, (Figure 4(a)), the easternmost network of surface rupture traces in the PVTR, is ~6.3 km long and has rupture strands spaced 3–180 m apart, with an average strike of 195° (+9°/−7°). Along the entire length of fault set 3, >90% of scarps in late Pleistocene deposits have been reactivated by Holocene faulting (scarp morphology 2).

We used cumulative right-lateral offset, vertical separation, and total offset at 288 displacement markers in Holocene to late Pleistocene units in the PVTR to determine the total magnitude of fault slip in different generations of alluvium (Figure 4, Figures 9-12) (Figures 4, 9, 10, 11, and 12). See Online Supplementary Table S12 for a full list of displacement markers. The magnitude of right-lateral field offsets ranges from 0.2–1.5 m. The magnitude of vertical separation measured from topographic profiles ranges from 0.0–0.6 m. Right-lateral and vertical offset magnitudes from backslipped DSMs range between 0.3–2.7 m and 0.0–0.8 m, respectively, with total oblique slip magnitudes ranging from 0.3–2.7 m. On individual scarps, total oblique slip magnitudes <0.5 typically occur along parallel, closely spaced (<1 m) rupture strands, or where slip is distributed between antithetic fault pairs defining small, meter-scale wide grabens. For scarps with total offsets >0.5 m, lateral to vertical slip ratios range between 3:1 and pure strike-slip, with a mean of ~8:1.

Furthermore, we can determine the number and relative timing of earthquakes on numerous scarps throughout the PVTR by combining our alluvial fan mapping with the magnitude of offset in subsequent generations of alluvium. Within the PVTR, no individual site contains a single, continuous surface rupture that offsets all generations of late Holocene alluvium. Below, we discuss offsets at four key rupture sites (RS 1–4; Figures 4, 9-12) that collectively establish the increase in cumulative offset by unit age and jointly reveal the late Holocene rupture history. At each of the rupture sites, we report the average total offset in each generation of alluvium and the average ratio of strike-slip to dip-slip for all offset alluvium at that site. The temporal rupture patterns at these four key rupture sites are consistently demonstrated at all of the rupture sites that we mapped within the PVTR. Furthermore, the increase in cumulative offset by unit age is supported by our COPD models, which are described at the end of this section.

Rupture site 1 (Figures 4(b), 9, and 10) contains a single, ~180-m-long surface rupture that offsets three different units, Qf8, Qf6a, and Qf6b, from youngest to oldest, respectively. Everywhere along the strike of rupture site 1, we measured no deformation in the youngest Holocene unit, Unit Qa, and we consistently measured smaller magnitudes of vertical separations and/or total offsets in Unit Qf8 compared to the relatively older Units Qf6b and Qf6a. At rupture site 1, the average total offset is 0.67 m (± 0.18 m) in Unit Qf8, 0.92 m (± 0.24 m) in Qf6a, and 1.27 m (± 0.31 m) in Qf6b, listed from youngest to oldest, respectively. Additionally, average strike-slip to dip-slip ratios at rupture site 1 are ~11:1. These data provide evidence that at rupture site 1: (a) Qf8 is the youngest unit with recent deformation, (b) Unit Qf8 has experienced less total slip and therefore fewer events than the older Units Qf6b and Qf6a deposits, and (c) Units Qf6b (older) and Qf6a (younger) deposits have similar cumulative displacements and may have experienced a similar number of events at rupture site 1.

Rupture site 2 (Figures 4(b), 9, and 11), which we subdivide into western rupture subset 2a and eastern rupture subset 2b, includes 6–8 surface ruptures that produce displacement in Unit Qf7 deposits and relatively older Unit Qf6a deposits. Everywhere along the strike of displacements in rupture site 2, we measured no deformation in younger, inset Units Qf8 and Qa deposits. Additionally, we consistently measured smaller magnitudes of cumulative offset in Unit Qf7 deposits compared to relatively older Unit Qf6a deposits. For example, within rupture subset 2a, we measured no displacement in Unit Qf7 deposits and average total offsets of 0.96 m (± 0.41 m) in Unit Qf6a deposits. Similarly, within rupture subset 2b, we measured smaller average total offsets of 0.69 m (± 0.16 m) in Unit Qf7 deposits, and larger average total offsets of 1.36 m (± 0.37 m) in Unit Qf6a deposits. The average strike-slip to dip-slip ratios at rupture subsets 2a and 2b are ~13:1 and ~23:1, respectively. These offset measurements and geomorphic relationships support the following interpretations at rupture site 2: (a) Unit Qf8 has not been locally deformed by an event, (b) Unit Qf7 deposits have experienced more events than the (locally) undeformed, inset Unit Qf8 deposits, (c) Unit Qf6a deposits have experienced greater total slip and more events than inset Unit Qf7 deposits, and (d) across all of rupture site 2, approximately 50% of the strands ruptured in an earthquake postdating the deposition of Unit Qf6a were then reactivated in an earthquake postdating the deposition of Unit Qf7.

At rupture site 3 (Figures 4(b), 9, and 11), several surface rupture strands produce displacement of Units Qf6b and Qf6a. Everywhere along the strike of displacements along rupture site 3, we measured no displacement of relatively younger, inset Units Qf7 and Qf8 deposits. Along a single ~0.4-km-long surface rupture, Units Qf6b and Qf6a are interfingered, and Unit Qf6b consistently has larger right-lateral offsets than the younger Unit Qf6a. Here, the average total offset is 1.11 m (± 0.16 m) in Unit Qf6a deposits and 2.02 m (± 0.24 m) in Unit Qf6b deposits. Similarly to rupture sites 1 and 2, the average strike-slip to dip-slip ratio at rupture site 3 is nearly pure strike-slip at ~33:1. These data support the following interpretations for rupture site 3: (a) Units Qf7 and Qf8 have been locally deformed by any Holocene earthquakes, (b) Unit Qf6a has more displacement than inset, (locally) undeformed, younger Units Qf7 and Qf8, and (c) Unit Q6a has a smaller magnitude of cumulative total displacement than relatively older Unit Qf6b deposits and has therefore experienced fewer earthquakes than Qf6b.

At rupture site 4, Figures 4(b), 9, and 12, several surface ruptures displace a Unit Qf7 terrace and a relatively older Unit Qf6a deposit. Here, two main surface rupture strands produce a NNE-trending graben that contains several, subparallel surface ruptures that cut through the down-dropped block (Figure 12(a)). Along the western graben-bounding fault (Figure 12(a); “W”), the average total slip of the Qf7-Qf6a unit boundary is 1.37 m (± 0.38 m), and average vertical separations of the Unit Qf7 terrace and the relatively older Unit Qf6a fan are 0.15 m (± 0.06 m), 0.40 m (± 0.06 m), respectively. Along the eastern graben-bounding fault (Figure 12(a), “E”), the average total slip of the Qf7-Qf6a unit boundary is 1.67 m (± 0.24), and average vertical separations are 0.42 m (± 0.06 m) in Unit Qf7 and 0.81 m (± 0.06 m) in Unit Qf6a. Additionally, two fault strands within the down-dropped graben produce minor vertical offsets of 0.06 m (± 0.08 m) in Unit Qf6a and are truncated by the inset Unit Qf7 terrace. Immediately east of the main graben, along a single fault strand, Unit Qf6a has total offsets of 1.62 m (± 0.2 m). Notably, the average strike-slip to dip-slip ratio at rupture site 4 is ~5:1, reflecting a much larger component of vertical displacement here, compared to rupture sites 1–3. Overall, these data support the following earthquake constraints for rupture site 4: (a) Unit Qf7 has smaller magnitudes of cumulative offset and has experienced fewer earthquakes than the relatively older Unit Qf6a, (b) cumulative magnitudes of slip in Qf7 are larger than single-event slip magnitudes in Qf8 at rupture site 1, and (c) each earthquake rupture of Units Qf6a and Qf7 produced a measurable component of vertical offset not observed in deposits at rupture sites 1–3.

To further evaluate the number and slip magnitude of earthquakes observed at rupture sites 1–4, we produced COPD models for field measurements of right-lateral offset and for LaDiCaoz backslipped reconstructions of right-lateral and total oblique slip, for each generation of offset late Holocene alluvium (Figure 13; Online Supplementary Figure S13(1)). These plots show several statistically significant peaks (α = 0.05) in the offsets of Qf7, Qf6a, and Qf6b, which we interpret as evidence for multiple earthquakes. We note that statistically significant peaks in COPD models could not be produced for the youngest generation of offset alluvium, Unit Qf8, due to the small number of reconstructed piercing points in this unit (n=1). The overlay of all statistically significant COPD models (Figure 13) consistently shows two peaks in Unit Qf7, and three peaks in Units Qf6a and Qf6b, with values of ~0.7 m, ~1.3 m, and ~2.2 m for peaks 1, 2, and 3, respectively (see Online Supplementary Table S13(1) for p-values). Notably, for each generation of offset alluvium, the modal values of lateral and total oblique offsets are comparable (Figure 13, Online Supplementary Table S13(1) and Figure S13(1)), which supports that right-lateral slip is the dominant mode of coseismic displacement in the PVTR, in accordance with our field observations.

There are two ways to interpret these COPD peaks. First, assuming that large earthquakes in the PVTR produce the same slip distribution in each subsequent event (i.e. a uniform slip model) [109, 110], the three sets of peaks in COPD plots imply preservation of three late Holocene earthquakes in the PVTR. Total displacements of ~0.7 m, ~1.3 m, and ~2.2 m would therefore be equivalent to geomorphic markers that were displaced in 1, 2, and 3 earthquake ruptures, respectively (Figure 13). While the uniform slip interpretation of the COPD models supports field-based evidence of three late Holocene earthquakes, this interpretation contradicts evidence for a fourth earthquake from field observations (e.g. rupture site 4; (Figures 4(b), 9(a), and 12). We also note that larger magnitude peaks (e.g. ~1.3 m and ~2.2 m) are not multiples of the smallest magnitude peak (~0.7 m), supporting our field observations that the magnitude of slip in the PVTR varies between events.

Instead, we pose an alternative interpretation that allows for variability in the magnitude of displacement in space and time (i.e. a variable slip model) [110, 111], an interpretation that is consistent with field observations at key rupture sites 1–4. Using COPD and field observations, we estimate a range of per-event displacement magnitudes of 0.6–1.1 m for late Holocene earthquakes in the PVTR. This range of displacement magnitudes is based on the following lines of evidence. First, peaks in COPD plots at offset magnitudes of ~0.7 m in Units Qf6b, Qf6a, and Qf7 overlap with the average total offset of 0.67 m (± 0.18 m) in Qf8 at rupture site 1, which is the youngest generation of late Holocene alluvium with displacement (Figures 9(a) and 10). Therefore, the most recent earthquake (MRE) likely produced total oblique offsets of ~0.7 m. Second, the difference between COPD peaks 1 and 2 in Qf7 (Figure 13) is ~0.7–0.8 m, which corresponds to average total oblique offsets of ~0.6–0.8 m in Qf7 at rupture subset 2b (figures 9(b) and 11). Moreover, at rupture subset 2b (Figure 11), rupture strands in Qf7 are truncated by younger, inset and undeformed Unit Qf8 deposits, signifying that the total oblique slip along these rupture strands is associated with displacement during earthquake 2 (i.e. the penultimate event). This supports that earthquake 2 likely produced average total oblique offsets of ~0.6–0.8 m, similar to total oblique offsets of ~0.7 m produced during the MRE. Finally, we constrain single-event offsets for earthquakes 3 and 4 using observations from rupture site 3 (figures 9(b) and 11), where average total offsets are ~2.0 m in Qf6b and ~1.1 m in Qf6a, and where younger Units Qf7 and Qf8 are not offset. These observations support three main interpretations: (1) A third earthquake is temporally bracketed by the deposition of Units Qf6a and Qf7, (2) a fourth earthquake is temporally bracketed by the deposition of Units Qf6b and Qf6a, and (3) the total oblique displacements associated with earthquakes 3 and 4 are ~1.1 m and ~0.9 m, respectively.

In summary, the inset, burial, and onlap relationships described in this section, as well as the differences in slip magnitudes in Units Qf6b, Qf6a, Qf7, and Qf8 (listed oldest to youngest) identified at rupture sites 1–4 and from COPD analyses, provide evidence that: (a) four earthquakes postdate the deposition of Qf6b, but no single surface rupture strand hosts all four earthquakes, (b) Qf6b and Qf6a bracket the timing of earthquake 4, (c) Qf6a and Qf7 bracket the timing of earthquake 3, (d) Qf7 and Qf8 bracket the timing of earthquake 2, and (e) Qf8 and the active wash (Qa) bracket the timing of earthquake 1, the MRE. Furthermore, combined interpretations from the COPD plots and single-event earthquake magnitudes observed at rupture sites 1–4, support that the PVTR is capable of producing single-event offset magnitudes of 0.6–1.1 m.

We combine our pIR-IRSL ages for offset Units Qf6b, Qf6a, Qf7, and Qf8 (Section 4.2 ), with key rupture-bracketing relationships determined from field mapping and COPD plots (Section 4.4.1 ), using a Monte Carlo Oxcal-like model [28] (Online Supplementary Method S14.1 and Supplementary Figure 14(1)) to estimate the timing of four late Holocene earthquakes in the PVTR. We report the following earthquake ages with 5% significance. We bracket the timing of the oldest event, earthquake 4, to 5.8–3.4 ka, which occurred after the deposition of Qf6b (5.9, 3.9 ka) and before the deposition of Qf6a (3.9, 3.3 ka). The next youngest event, earthquake 3, occurred between 3.8 and 2.2 ka, after the deposition of Qf6a (3.9, 3.3 ka) and before the deposition of Qf7 (2.6, 2.2 ka). Earthquake 2, the penultimate event, occurred between 2.4 and 0.6 ka, after the deposition of Qf7 (2.6, 2.2 ka) and before the deposition of Qf8 (~2.2–0.5 ka). Earthquake 1 (the MRE) occurred after the deposition of Qf8 and before the deposition of the active wash (Qa). To evaluate the timing of earthquake 1, we use the age of Unit Qf8 as the maximum bound, by correlating Qf8 surfaces in the PVTR to morphologically similar surfaces along the western Panamint Range front that are dated to ~0.8–0.5 ka [48, 50, 51]. To estimate the minimum age of earthquake 1, we infer that earthquake 1 occurred prior to the beginning of the historical records in the Death Valley region, equivalent to the date of the earliest recorded earthquake, the 1872 Owens Valley earthquake, around ~150 ybp. We use these bounding ages to estimate the timing of the MRE to 640–160 ybp, or ~0.64–0.16 ka. Finally, we estimate late Holocene slip rates in the PVTR using three constraints: (1) minimum and maximum values of slip-per event (~0.6–1.1 m, see Section 4.4.1), (2) cumulative slip ranging between 2.4 and 4.4 m, estimated by multiplying slip-per-event values by four earthquakes, and (3) the maximum and minimum age of Qf6b (5.9, 3.9 ka, see Section 4.2), which provides an upper bound for the age of the oldest event, earthquake 4. From these constraints, we late Holocene slip rates of 0.9–2.5 mm/yr for the PVTR.

We define paleo-surface rupture maps for each of the four late Holocene earthquakes preserved in the PVTR (Figure 14), using our field observations to determine which alluvial generations bracket earthquake rupture timing (Section 4.4.1). Earthquake 4 ruptures are mapped where Qf6b is offset and Qf6a is not, or where cumulative offset in Qf6b is greater than cumulative offset in Qf6a. Earthquake 3 ruptures are mapped where Qf6a is offset and Qf7 is not, or where cumulative offset in Qf6a is greater than cumulative offset in Qf7. Earthquake 2 ruptures are mapped where Qf7 is offset, and where either Qf8 is not offset or there is larger total offset in Qf7 than Qf8 along strike. Earthquake 1 (MRE) ruptures are mapped where surface ruptures offset Q8 and are truncated by Qa, or where the magnitude of slip in Qf7 is consistent with more than one event (Figure 13). Potential ruptures for each event are mapped in locations where rupture is permissible but not well-constrained due to the absence of alluvial units that immediately predate or postdate the event.

From our paleo-rupture maps (Figure 14), we observe distinct variations in the location, total number, and total length of surface ruptures accommodating coseismic slip in each of the four late Holocene earthquakes (Figure 14; Online Supplementary Figure S15). Earthquake 4, the oldest Holocene earthquake identified, appears to have ruptured the entire PVTR, extending ~9 km between the northernmost and southernmost mapped surface ruptures and displacing fault strands along the entire ~3.5 km width of the PVTR (Figure 14). Similarly, earthquake 3 ruptured a total length of 6–9 km, including the entirety of the eastern and central rupture strands, and possibly the most west to northwestern surface rupture strands (Figure 14). In contrast, the two youngest events, earthquake 2 and earthquake 1 (the MRE), appear to have only ruptured a subset of faults in the PVTR. Earthquake 2 extended along a smaller subset of surface ruptures in the western and central sections of the PVTR, producing a total rupture length of 6–8 km. The abundance and wide spatial distribution of Qf7 deposits that bracket this rupture suggest these total rupture lengths are robust. Earthquake 1, the MRE, only ruptures a 4- to 5-km-long subset of dominantly north-south trending surface ruptures in the central PVTR. However, we recognize that the preservation of young scarps in poorly consolidated Qf8 may be limited, such that the spatial extent of MRE ruptures may be larger (Figure 14).

Additionally, our paleo-rupture maps show distinct trends in the strikes and locations of surface rupture strands activated in each late Holocene earthquake (Figure 14). In general, the central zone of surface rupture strands (Figures 4 and 14, and Online Supplementary Figure S15) exhibited rupture in all four late Holocene earthquakes. Notably, the faults in the central PVTR are the most geometrically complex and contain the widest range of surface rupture strikes (SSW- to SSE-striking, Section 4.3.2 ). In contrast, the easternmost and westernmost surface rupture strands, where faults are dominantly NNE-striking, only appear to have accommodated slip during the older events, earthquakes 3 and 4 (upper panels in Figure 14). These observations suggest that rupture preferentially occurs along the most geometrically complex, central PVTR faults, whereas coseismic displacement of the generally NE-striking, westernmost and easternmost faults occurs less frequently. We also find that nearly all the late Pleistocene scarps along the central and eastern extents of the PVTR have been reactivated by late Holocene ruptures (Section 4.3.2 ), while the majority of Pleistocene scarps along the western to southwestern boundary of the PVTR are abandoned (Figure 14).

Our high-resolution tectono-geomorphic mapping shows that the PVTR is a SSW-striking, ~5-km-wide, ~9-km-long system of faults located in the stepover between the southern Ash Hill fault and the Panamint Valley fault. The PVTR consists of dozens of distributed, parallel, and en échelon, active surface rupture strands that collectively accommodate a combination of right-lateral and oblique-normal slip. PVTR surface ruptures strike ~190° on average (Figures 1(b), 4, and 14) and are ~30° oblique to the strikes of the Ash Hill (~160–165°) [28, 55], and central to southern Panamint Valley (~150–170°) faults (e.g. [27, 38-40, 46-51]). While we identify four late Holocene earthquakes in the PVTR with recurrence intervals of ~1.0–1.5 kyr, we also identify strong spatial and temporal variations in the combinations of surface rupture strands that link up to accommodate coseismic strain release (Figure 14). These spatiotemporal variations of seismic rupture patterns in the PVTR, in addition to the geometry of surface ruptures and a close proximity to nearby faults, suggest that it may facilitate strain transfer between the Ash Hill and Panamint Valley faults. Though these three faults do not directly merge at the surface, the mapped fault geometries, range of strike-slip to oblique-slip kinematics, and paleo-rupture map patterns that we present in this study provide evidence that the PVTR may act as a structural link between these adjacent faults in the subsurface. In particular, the most northeastern PVTR scarps directly project into a network of SSW-striking Holocene scarps along the central Panamint Valley fault (northeast of Ballarat), while the most northwestern PVTR scarps have a distinct northwest-stepping trend and ultimately project into Holocene scarps on the southern Ash Hill fault at a latitude of 36° 3’N (Figure 1 and (b), 4, 14, and Online Supplemental Plate S10).

Furthermore, our data provide constraints for how strain is partitioned between active faults in the Panamint Valley during the late Holocene. First, our measured cumulative displacements demonstrate that faults in the PVTR have similar average ratios of strike-slip to dip-slip (8:1) as both the Ash Hill fault (~6.5:1) [28] and the southernmost Panamint Valley fault (~5:1) [38, 51]. Second, we find that while the PVTR has similar fault slip kinematics as both the southern Panamint and Ash Hill faults, the per-event slip magnitudes and late Holocene slip rates on PVTR faults (~0.6–1.1 m of slip per event and a ~0.9–2.5 mm/yr slip rate) are more comparable to those reported on the Ash Hill fault (1.0 ± 0.2 m of slip per event and a ~0.6–1.4 mm/yr slip rate) [28]. In comparison, the per-event slip magnitudes and late Holocene slip rates of the PVTR and Ash Hill faults are 2 to 8 times smaller than those reported for the Panamint Valley fault (2.5, 5 m of slip-per event and a 1.5–6.0 mm/yr slip rate) [28, 47-51, 55] and 3 to 12 times larger than the slip rates along the western bounds of the Slate Range (~0.17–0.35 mm/yr) [60], southwest of the PVTR. Additionally, at 36°N, where the southern PVTR and central Panamint Valley fault begin to overlap (Figures 1(a), and 1(b), 14 and Online Supplementary Material S10), prior data have posited that the Panamint Valley fault may undergo a south to north transition in slip kinematics [38, 39, 47-51]. North of the PVTR, palinspastic reconstructions of displacement along the central Panamint Valley fault show that it accommodates a larger proportion of extension compared to the Panamint Valley fault south of the PVTR [38, 39]. This along-strike change in slip kinematics of the Panamint Valley fault may reflect changes in strain accommodation and partitioning related to the intersection of the Ash Hill-PVTR-Panamint Valley faults. We postulate that the PVTR may help to partition transtensional strain between the Panamint Valley and Ash Hill faults, similar to strain partitioning that has been well-documented in transpressional systems [112-115].

There are several existing models of subsurface fault geometry in the Panamint Valley, each with separate implications for how slip could be transferred between the Ash Hill and Panamint Valley faults at depth. One end-member model, the SCEC Model (Figure 15, Model A) [52, 53], assumes the Panamint Valley fault is a steeply (70°–90°) west-dipping structure along its entire ~100 km length. While the SCEC Model is one viable geometric representation of the Ash Hill and southern Panamint Valley faults, it does not include the geometrically complex faults that we map in this study in the stepover between the Ash Hill and Panamint Valley faults. In addition, the high-angle geometry and strike-slip kinematics of the central and northern segments of the Panamint Valley fault in the SCEC Model appear to conflict with evidence that the low-angle Panamint Valley detachment may still be active at depth [38-40, 45, 50].

A second end-member geometric model, proposed by Walker et al. [38] (Figure 15, Model B), proposes that the central Panamint Valley fault is instead a low-angle fault and is physically connected to the Manly Pass-Searles Valley fault zone in the subsurface. This model suggests that strain partitioning occurs between the Panamint-Manly Pass-Searles Valley faults and the Ash Hill-Manly Pass-Searles Valley faults [38]. In this model, the dip of the Panamint Valley fault shallows north of ~36°N, compatible with data that suggest that the low-angle Panamint Valley detachment is still active [38-40, 45, 50]. However, the Walker Model [38] contains stepover fault geometries that are incompatible with the results of our mapping in the PVTR. In the Walker Model, the southern tip of the Ash Hill fault extends to the latitude of the PVTR (Figure 15), with S to SE-striking strands intersecting the northern trace of the Manly Pass fault [38]. During our high-resolution mapping, we found no evidence for SE-striking fault strands in Holocene alluvium within the PVTR, nor evidence for reactivation of Pleistocene scarps along the southwestern edge of the PVTR (Figures 1(b), 4(a), and 14). Furthermore, this model appears to conflict with the observed geometry of high-angle faults at the base of the Panamint Range [54]

We propose a third, hybrid model (Figure 15) that allows for strain in Panamint Valley to be partitioned between a deeper, low-angle, remnant detachment and shallow high-angle faults, including SSW-striking faults that are compatible with our mapping of the PVTR, in the stepover between the Ash Hill and Panamint Valley faults. This model includes a low-angle Panamint-Manly Pass-Searles Valley detachment (Figure 15, Model C: “2d,” inactive at the surface), high-angle faults at the base of the Panamint Range north of the PVTR (Figure 15, Model C: “2n”), high-angle Ash Hill (Figure 15, Model C: “1”) and southern Panamint Valley faults (Figure 15, Model C: “2s”), and a small number of high-angle PVTR faults that may merge with the Panamint detachment at depth (Figure 15, Model C: “4”). This Hybrid Model is an intermediate model between the SCEC and Walker Models and appears to be compatible with all currently available fault geometry data. This proposed partitioning between deep and shallow structures in Panamint Valley is similar to that observed in regions with active low-angle normal faults, such as those in central and southern Italy [116, 117].

To a first order, the fault geometries depicted in our Hybrid Model (Figure 15) imply that the PVTR either acts as a fault tip for the Ash Hill fault or accommodates transtensional deformation within a strike-slip stepover between the southern Panamint Valley and Ash Hill faults. However, the SSW-striking (182°–195°) PVTR fault geometries in the central Panamint Valley are inconsistent with typical dextral strike-slip fault tip geometries (e.g. en échelon or horsetail splays that would be located southwest of the southern tip of the Ash Hill fault), and the extensional displacement documented within the PVTR is inconsistent with the transpression that would be predicted to occur in a left stepover between the right-lateral, oblique-normal slip of the southern Panamint and Ash Hill faults. Instead, the SSW-striking (~190°) PVTR faults appear to accommodate net transtensional displacement, in the stepover between the SSE-striking Ash Hill (~160-165°) and Panamint Valley (~150–170°) faults. These PVTR slip kinematics appear to be most compatible with the Hybrid Model (Figure 15, Model “C”) that invokes a gently west-dipping (~20°) detachment projecting under the PVTR and Ash Hill faults. For example, the geometry of PVTR surface ruptures may be analogous to accommodation faults that create secondary horst and graben complexes in the hanging wall of a major breakaway fault [118, 119], potentially reflecting reactivation of a remnant low-angle Panamint detachment at depth. If the PVTR and/or Ash Hill faults directly merge with the Panamint Valley fault at depth, this would provide a physical link for strain transfer and a gently dipping fault to accommodate the normal extension of the regional transtensional slip vector.

Our results further indicate that strain transfer between faults in the Panamint Valley occurs via slip during earthquakes and that the PVTR may co-rupture with other faults in the Panamint Valley. There are three lines of evidence that support this interpretation. First, displacement length scaling relationships [120-122] estimate that single-event offsets in the PVTR ranging between 0.6 and 1.1 m, as indicated by field observations and COPD models (Figure 13), should be associated with total rupture lengths of ~20–30 km. However, rupture lengths within the PVTR for the four documented late Holocene earthquakes are only 4–9 km (Figure 14). As cumulative displacement magnitudes do not generally appear to die out along either the northern or southern extent of the PVTR, these scaling relationships suggest two things: (1) the PVTR does not rupture independently of nearby faults, but likely co-ruptures with either the Ash Hill and/or the Panamint Valley faults and (2) there may be additional fault segments that are obscured by playa deposits in the ~3 km stepover between the eastern extent of our mapped PVTR surface ruptures and the Panamint Valley fault (Online Supplementary Material S10). Second, the similarity in timing of the last 3–4 earthquakes on the Panamint Valley, Ash Hill, and PVTR faults (Figure 16; Online Supplemental Figure S16(1)) implies that the PVTR may co-rupture with the Ash Hill and/or the Panamint Valley faults. This is further supported by the observation that late Holocene rupture-bracketing alluvial units along the PVTR, Ash Hill, and Panamint Valley faults have similar morphologies (Figure 16; On;line Supplementary Material S10, Online Supplementary Figure S14(1) and S16(1)), and by the similarities in the recurrence intervals of earthquakes on the PVTR and Ash Hill faults (1, 1.5 kyr). Finally, we find that late Holocene earthquakes in the PVTR have similar single-event slip magnitudes and transtensional slip kinematics as the Ash Hill fault. These data strongly suggest that the PVTR co-ruptures with the southern Ash Hill fault.

Furthermore, our data also provide evidence that the PVTR may transfer seismogenic strain between the Ash Hill and Panamint Valley faults. Similarities in the morphology of alluvial units that bound surface ruptures, number of late Holocene earthquakes, and the overlap in the timing of earthquakes on the PVTR, Ash Hill, and Panamint Valley faults (Figure 16, Online Supplementary Figure S16(1)) indicate that all three faults could rupture in the same or related events. We also note that the minimum inter-fault spacings between the PVTR-Ash Hill and PVTR-Panamint Valley faults (~2 km and ~3 km, respectively) are comparable to fault spacings where seismogenic strain can be transferred between adjacent systems, as observed from historical and paleoseismic ruptures [18-26], and from static and dynamic rupture simulations [14, 123-125]. Collectively, these data permit the occurrence of multifault earthquake triggering between the PVTR, Ash Hill, and Panamint Valley faults during one or more earthquakes during the late Holocene.

Our proposed Hybrid Model C (Figure 15) provides one example of fault geometries that may permit seismogenic strain transfer. Seismogenic strain transfer between the Panamint Valley, PVTR, and Ash Hill faults may be facilitated by slip along a low-angle detachment that physically links these three fault zones (Figure 15, Model C: Hybrid Model). This scenario is analogous to the kinematic coupling observed between hanging wall faults and a basal detachment in Death Valley [126]. In this case, the low-angle detachment may help partition strike-slip and dip-slip between the Panamint Valley and Ash Hill faults through strain release during seismogenic ruptures.

Several lines of evidence, including surface rupture mapping, rupture-bracketing late Holocene alluvial units, and paleo-rupture maps, demonstrate that the path that coseismic slip may take during an earthquake rupture, through the PVTR, varies over multiple earthquake cycles (Figure 14). This variability occurs in both small, meter-scale variations in which rupture strands are accommodating coseismic slip (Figure 11; rupture site 2), and by large, kilometer-scale variations in the persistence of paleo-surface rupture networks that we identify in the PVTR (Figure 14). Three key patterns emerge from our detailed tectono-geomorphic paleo-rupture map data.

First, these maps show variability in the total area ruptured during four earthquakes. For example, the older events, earthquakes 4 and 3, appear to have ruptured the entire length and width of the PVTR, while the more recent events, earthquakes 2 and 1, ruptured a smaller subset of surface ruptures localized in the central PVTR (Section 4.4.1, Figure 13). Interestingly, earthquakes 4 and 3 produced a greater total length (~6–9 km) of surface ruptures, and larger single-event offsets of 0.9–1.1 m, compared to relatively shorter total length (~4–8 km) of surface ruptures and relatively smaller single-event offsets of ~0.6–0.8 m estimated for earthquakes 2 and 1. Such a correlation may suggest that earthquakes with larger moment magnitude and larger fault slip may rupture more strands within the PVTR and provide a higher likelihood of ruptures jumping to adjacent fault strands than lower magnitude events.

Second, our paleo-surface rupture maps (Figure 14) also suggest that there may be a geometric preference for which scarps are repeatedly ruptured in subsequent earthquakes. For example, many of the surface ruptures activated during the older events, earthquakes 4 and 3, along the westernmost and easternmost extents of the PVTR, are SSW-striking (Figure 14). These same SSW-striking surface ruptures do not appear to have ruptured during the more recent earthquakes 2 and 1, where most of the slip appears to have localized on N to S-striking strands in the central PVTR (Figure 14).

Third, the variations in the geometry and spacing of faults accommodating coseismic rupture in the PVTR imply that the “jump distance” (i.e. the distance that an earthquake jumps between faults) that accommodates dynamic rupture propagation [14, 123-125, 127, 128] varies with each earthquake. Our paleo-rupture maps (Figure 14) suggest that the minimum structural gaps between the Ash Hill and PVTR faults are ~2–3 km, ~2–3 km, ~3–5 km, and ~5–6 km for earthquakes 4, 3, 2, and 1, respectively. Similarly, the minimum structural gaps between the Panamint Valley and PVTR faults for each of the four earthquakes are ~3 km, ~3–4 km, ~3–4 km, and ~5–6 km for earthquakes 4, 3, 2, and 1, respectively. These data imply that the spatiotemporal variations in strain accumulation and release in a geometrically complex fault stepover, such as the PVTR, may control the jump distance required for dynamic rupture between adjacent faults. Ultimately, the PVTR faults decrease the stepover distance between the nearby Ash Hill, Panamint Valley, and Manly Pass-Searles Valley fault zones, increasing the likelihood that dynamic rupture can jump between these adjacent faults [14, 123-125]. However, the distance that a dynamic rupture must breach across the PVTR to potentially generate a multifault earthquake appears to depend on which PVTR fault segments are accommodating seismogenic strain.

Our results demonstrate the utility of high-resolution tectono-geomorphic mapping to quantify the temporal and spatial heterogeneities of paleoseismic rupture geometry and slip kinematics in zones of complex and distributed fault geometries. Paleo-rupture maps are particularly useful in assessing whether paleo-rupture paths may be the result of multifault, dynamically or statically triggered rupture, similar to multifault earthquakes that have been documented by historical and paleoseismic ruptures in the southwestern U.S. [18-26]. In this study, we provide one example of spatiotemporal variations in paleo-rupture patterns during the late Holocene, which may indicate that strain release in Panamint Valley occurs via seismogenic strain transfer between adjacent faults. The fault geometries we document in this study are comparable to fault geometries observed in stepovers that accommodated the 1992 Landers multifault rupture [129, 130], and in broadly distributed zones of deformation observed in the 2020, Mw 6.5 Monte Cristo earthquake (central Walker Lane), southwestern Nevada [131]. Temporally and spatially dependent fault conditions, in addition to transient stress perturbations, appear to influence the likelihood of seismicity and/or dynamically triggered events [14, 123-125, 127, 128]. Therefore, past rupture complexity, as documented by variations in paleo-rupture surface networks over subsequent earthquake cycles, is relevant in determining how strain is accommodated in structurally complex fault discontinuities.

Spatiotemporal heterogeneity in rupture paths, like that documented within the PVTR, provides important boundary conditions for rupture models simulating the transfer or arrest of rupture through structurally complex transfer zones. Our data suggest that if seismogenic slip is transferred between the Ash Hill and Panamint Valley faults through the PVTR, there are many combinations of fault segments and/or surface ruptures that may accommodate this transfer. These outcomes emphasize that the dynamic linking of separate, but adjacent faults through geometrically complex fault boundary zones is an important process that should be considered in dynamic rupture models and earthquake hazard simulations. Identifying regions of geometrically complex faulting and regions where there may be local perturbations of the stress field is important when assessing which faults may be critically stressed or at risk of failure. Paleo-rupture maps of well-preserved fault discontinuities may help us identify whether specific zones have been capable of propagating multifault ruptures in the past. Classifying the geometric signatures of multifault rupture from paleo-rupture maps may provide constraints for modeling to identify what boundary conditions are required for multifault earthquake triggering, or conditional dynamic rupture.

In this study, we combine high-resolution tectono-geomorphic mapping and feldspar pIR-IRSL dating of late Holocene alluvium with fault slip kinematics in the PVTR to generate paleo-rupture maps that demonstrate how the release of seismogenic strain changes in space and time over multiple earthquake cycles. We constrain the timing of four late Holocene earthquakes in the PVTR and demonstrate that they have similar timing, recurrence intervals, slip magnitudes, and slip kinematics as the adjacent Ash Hill and Panamint Valley faults. These data suggest that seismogenic strain transfer may have occurred between the Ash Hill and Panamint Valley faults, assisted by the geometrically complex and distributed PVTR fault zone. Additionally, our data provide constraints for the likely subsurface fault geometries in the Panamint Valley that may accommodate this strain transfer. We suggest that a deeper, remnant segment of the low-angle Panamint detachment may help to accommodate normal-oblique slip and transfer strain from the southern Panamint Valley fault to the Ash Hill fault via the PVTR. Furthermore, our paleo-rupture maps of late Holocene PVTR earthquakes provide evidence that the combination of fault planes that are activated, and the jump distance between faults that aid in seismogenic strain transfer varies over multiple earthquake cycles. The geometric heterogeneity that we present in this study must be compatible with any model that either transfers or inhibits rupture through this type of structurally complex zone. Our work highlights the utility of tectono-geomorphic paleo-rupture mapping for identifying the paths that earthquake ruptures take between adjacent faults and for characterizing spatiotemporal variations of earthquake rupture in the stepovers between faults over multiple earthquake cycles.

Additional data supporting the results and interpretations in this article are presented in the online Supplemental Materials. Digital surface models (DSMs) generated in this study are available via the Open Topography portal (https://doi.org/10.5069/G9PC30M4). Our database of measured fault offsets, and codes used to reconstruct vertical separation (e.g. reconstructTopo), is publicly available from Dryad (https://doi.org/10.5061/dryad.dr7sqvb6j). The luminescence data release can be obtained from the U.S. Geological Survey ScienceBase Catalog (https://doi.org/10.5066/P9P81WDF). Lidar DEMs used in this study are based on services provided to the Plate Boundary Observatory by NCALM (http://www.ncalm.org), operated by UNAVCO for EarthScope (http://www.earthscope.org), and supported by the National Science Foundation (Grants EAR-0350028 and EAR-0732947).

The authors declare that there are no conflicts of interest regarding the publication of this paper.

This project has been supported in part by: the National Cooperative Geologic Mapping Program Award No. G21AC10725-00, the GSA AGeS2 program via the National Science Foundation Grant Nos. EAR-1759200 and EAR-1759353, and the NAU Duebendorfer-Barnes Endowment (LaPlante and Regalla); the GSA Student grants program and the NAU David Sanders Geology Research Assistantship Award (LaPlante).

We thank the following students who aided in field work, data processing, and synthesis: Letty Rodriguez, Emma Foley, Heather Elliott, Amanda Binkley. We thank Emma Krolczyk for her help with sample processing at the U.S. Geological Survey Luminescence Laboratory. We thank Andrew Cyr, Eric Kirby, and the late Paul Umhoefer for their help with local and regional field mapping interpretations. Any use of trade, firm, or product names is for descriptive purposes only and does not imply endorsement by the U.S. Government.

Supplemental Table S1 is a table of terminology used for alluvial fan weathering descriptions. Supplemental Table S2 is a table of terminology used for desert pavement descriptions. Supplemental Figure S3 contains shaded relief models showing the difference in resolution of imagery (e.g. digital elevation models, digital surface models, lidar) used for mapping in this study. Supplemental Figure S4 contains seven examples of aerial imagery, digital surface models (DSMs), and DSM derivatives generated using drone-based structure from motion, which we used to map surface ruptures in this study. Supplemental Methods S5 describes our methods for reconstructing vertical separation of offset alluvial fans using our reconstructTopo code (available from Dryad https://doi.org/10.5061/dryad.dr7sqvb6j). Supplemental Methods S6 outlines our methods to combine the uncertainties for backslipped reconstructions of a single piercing line. Supplemental Methods S7 describes our methods of filtering offsets with high uncertainty (Methods S7.1), examples of offset uncertainty for measured geomorphic features (Figure S7.1), and methods of optimization (Methods S7.2) to produce statistically significant cumulative offset probability distribution (COPD) models. Supplemental Methods S8 describes our methods of preparing samples for pIR-IRSL dating (S8.1), and methods for calculating the equivalent dose D_e (Methods S8.2) and dose rate D_r (Methods S8.3) for K-feldspar. Supplemental Tables S9 are comprised of three specification tables for pIR-IRSL sample processing including: (a) aliquot filtering acceptance criteria (Table S9.1), (b) single aliquot regeneration protocol (Table S9.2), and (c) machine parameters used for luminescence analyses (Table S9.3). Supplemental Plate S10 includes a preliminary surficial map of central Panamint Valley, including a full EDMAP plate, the list of map units (LMU), correlation of map units (CMU), and four additional inset maps showing other examples of densely spaced, diffuse, and complex faulting in late Holocene alluvium. This supplementary map plate can be found in a separate .pdf file. Supplemental Figure S11 shows Figure 8 without annotations. Supplemental Table S12 includes a table of all measured field, backslipped, and vertically reconstructed offset measurements. This supplementary table can be found in a separate .csv file. Supplemental Results S13 includes a table of model parameter results (S13.1) and a figure of cumulative offset probability distribution (COPD) curves (S13.2) for optimization models produced using methods described in Supplemental Methods S7.2. Supplemental S14 contains the methods (Method S14.1) and results (Figure S14.1) of the Monte Carlo Oxcal-like age model that we used to determine the 95% confidence interval for earthquake timing for 4 late Holocene events in the PVTR. Supplemental Figure S15 shows 12 examples of paleo-rupture maps produced in this study, including the four maps from the main text (Figure 14), with different qualitative spatial bins of 0.04 km², 0.16 km², and 0.36 km². Supplemental 16 (S16) contains methods for a Monte Carlo simulation of temporal earthquake overlap (Methods S16.1) and model results (Figure S16.1) used to determine the likelihood that individual earthquakes on the PVTR, Ash Hill and/or Panamint Valley faults ruptured within a specific temporal window.